Decomission compute node


I am in the process of decommissioning a compute node in Openstack.
I cannot find an official way of removing a compute node from Openstack with Juju or manually with the help of documentation.

What i did find is helpful documentation from Openstack users at home or other companies and their own Openstack solution (RedHat, Mirantis).

Why hasn’t a way to easily remove a compute node been developed for Juju?

Hey @emiljacero,

It would indeed be handy with explicit lifecycle docs in the deployment guide.

Have you tried removing the compute node in Juju already? If so, where did you get stuck?

Last time I removed a compute node I just ran juju remove-unit nova-compute/x, but I redeployed it with the same name after recommissioning in MAAS so I unfortunately do not remember if the charm properly removed the service agents or not. If not, you should be able to remove them with the OpenStack CLI somewhat similar to Red Hat docs for removing compute nodes.

Regarding docs, @pmatulis, are there any plans for such lifecycle docs somewhere? If not, would it be welcome to send some suggestions to the deployment guide to get started?

Hi @szeestraten,

I ran juju remove-unit nova-compute/x and then juju remove-machine x.
That worked as expected. It however only removed the unit and machine. Those commands didn’t trigger any cleanup in the database or anything.

I don’t really know if Juju is able to observe that and react to it, and if it was able to, how to do it securely and reliably.

For now i will be following the manual steps i found around the web :slight_smile:


I see, that’s what I was afraid of. I took a look at the bug list for the nova-compute charm and I found: Bug #1317560 “remove-unit of nova-compute does not actually remo...” : Bugs : OpenStack nova-compute charm which seems to be relevant one for this use case.

P.S. There is a tag for bugs related to removing units or scaling down with OpenStack charms called scaleback which might be interesting for you.